R Bootcamp, Module 5:
Data Manipulation Using the tidyverse

Created by: Kellie Ottoboni, Rochelle Terman and Chris Krogslun (UC Berkeley)
Edited by: John S. Erickson (The Rensselaer IDEA)

Updated 17 Feb 2021 (JSE)

Philosophy

It is often said that 80% of data analysis is spent on the process of cleaning and preparing the data. (Dasu and Johnson, 2003)

Thus before you can even start on any sort of sophisticated analysis or plotting, you first must:

  1. Manipulate data frames: filtering, summarizing, and conducting calculations across groups
  2. Tidy data into the appropriate format

Historically there are two schools of thought within the R community…and at RPI!

Welcome to the tidyverse!

This tutorial shows you some of the tidyverse tools so you can make an informed decision about whether you want to continue to suffer through Base R or enter the tidyverse.

tidyverse Packages

See also: https://www.tidyverse.org/packages/

Check out the tidyverse cheatsheet!

See also: https://github.com/rstudio/cheatsheets/raw/master/data-transformation.pdf (Updated 17 Jan 2021)

Example Dataset

For this unit, we’ll be working with the “Gapminder” dataset, which is excerpt of the data available at Gapminder.org. For each of 142 countries, the data provides values for life expectancy, GDP per capita, and population, every five years, from 1952 to 2007.

gapminder <- read.csv("../data/gapminder-FiveYearData.csv",
          stringsAsFactors = TRUE)
head(gapminder)
##       country year      pop continent lifeExp gdpPercap
## 1 Afghanistan 1952  8425333      Asia  28.801  779.4453
## 2 Afghanistan 1957  9240934      Asia  30.332  820.8530
## 3 Afghanistan 1962 10267083      Asia  31.997  853.1007
## 4 Afghanistan 1967 11537966      Asia  34.020  836.1971
## 5 Afghanistan 1972 13079460      Asia  36.088  739.9811
## 6 Afghanistan 1977 14880372      Asia  38.438  786.1134

Another, more complete way to sanity-check our data types:

str(gapminder)
## 'data.frame':    1704 obs. of  6 variables:
##  $ country  : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
##  $ pop      : num  8425333 9240934 10267083 11537966 13079460 ...
##  $ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
##  $ gdpPercap: num  779 821 853 836 740 ...

Dataframe Manipulation using Base R Functions

So far, you’ve seen the basics of manipulating data frames, e.g. subsetting, merging, and basic calculations. For instance, we can use base R functions to calculate summary statistics across groups of observations, e.g. the mean GDP per capita within each region:

mean(gapminder[gapminder$continent == "Africa", "gdpPercap"])
## [1] 2193.755
mean(gapminder[gapminder$continent == "Americas", "gdpPercap"])
## [1] 7136.11
mean(gapminder[gapminder$continent == "Asia", "gdpPercap"])
## [1] 7902.15

But this isn’t ideal because it involves a fair bit of repetition. Repeating yourself will cost you time, both now and later, and potentially introduce some nasty bugs.

We want our code to more closely resemble our thinking, which is “get the means by continent

Dataframe Manipulation Using dplyr

Luckily, the dplyr package provides a number of very useful functions for manipulating dataframes. These functions will save you time by reducing repetition. As an added bonus, you might even find the dplyr grammar easier to read.

Here we’re going to cover six of the most commonly used functions as well as using pipes (%>%) to combine them.

  1. select()
  2. filter()
  3. group_by()
  4. summarize()
  5. mutate()
  6. arrange()

dplyr will be installed when installing tidyverse, or you can install it by itself:

# not run
# install.packages('dplyr')

Now let’s load the package:

library(dplyr)

dplyr select()

Imagine that we just received the gapminder dataset, but are only interested in a few variables in it. We could use the select() function to keep only the columns corresponding to variables we select.

year_country_gdp <- select(gapminder, year, country, gdpPercap)
head(year_country_gdp)
##   year     country gdpPercap
## 1 1952 Afghanistan  779.4453
## 2 1957 Afghanistan  820.8530
## 3 1962 Afghanistan  853.1007
## 4 1967 Afghanistan  836.1971
## 5 1972 Afghanistan  739.9811
## 6 1977 Afghanistan  786.1134

If we open up year_country_gdp, we’ll see that it only contains the year, country and gdpPercap. This is equivalent to the base R subsetting function:

year_country_gdp <- gapminder[,c("year", "country", "gdpPercap")]
head(year_country_gdp)
##   year     country gdpPercap
## 1 1952 Afghanistan  779.4453
## 2 1957 Afghanistan  820.8530
## 3 1962 Afghanistan  853.1007
## 4 1967 Afghanistan  836.1971
## 5 1972 Afghanistan  739.9811
## 6 1977 Afghanistan  786.1134

But, as we will see, dplyr makes for much more readible, efficient code because of its pipe operator.

dplyr pipes

Above, we used what’s called ‘normal’ grammar, but the strengths of dplyr lie in combining several functions using pipes.

Pipes take the input on the left side of the %>% symbol and pass it in as the first argument to the function on the right side.

Since the pipe grammar is unlike anything we’ve seen in R before, let’s repeat what we’ve done above using pipes.

year_country_gdp <- gapminder %>% select(year, country, gdpPercap)

Fun Fact: There is a good chance you have encountered pipes before in the Linux shell. In R, a pipe symbol is %>% while in the shell it is |. But the concept is the same!

dplyr filter()

Now let’s say we’re only interested in African countries. We can combine select() and filter() to select only the observations where continent is Africa.

year_country_gdp_africa <- gapminder %>%
    filter(continent == "Africa") %>%
    select(year,country,gdpPercap)

First we pass the gapminder dataframe to the filter() function, then we pass the filtered version of the gapminder dataframe to the select() function.

Both the select() and filter() functions subset the data frame. The difference is that select() extracts certain columns, while filter() extracts certain rows.

Note: The order of operations is very important in this case. If we used select() first, filter() would not be able to find the variable continent since we would have removed it in the previous step.

dplyr Calculations across Groups

A common task you’ll encounter when working with data is running calculations on different groups within the data. For example, what if we wanted to calculate the mean GDP per capita for each continent?

In base R, you would run the mean() function for each subset of data:

mean(gapminder$gdpPercap[gapminder$continent == "Africa"])
## [1] 2193.755
mean(gapminder$gdpPercap[gapminder$continent == "Americas"])
## [1] 7136.11
mean(gapminder$gdpPercap[gapminder$continent == "Asia"])
## [1] 7902.15
mean(gapminder$gdpPercap[gapminder$continent == "Europe"])
## [1] 14469.48
mean(gapminder$gdpPercap[gapminder$continent == "Oceania"])
## [1] 18621.61

That’s a lot of repetition! To make matters worse, what if we wanted to then add these values to our original data frame as a new column? We would have to write something like this:

gapminder$mean.continent.GDP <- NA  # Initialize a new column

# Write the values into the new column, for each continent 
gapminder$mean.continent.GDP[gapminder$continent == "Africa"] <- mean(gapminder$gdpPercap[gapminder$continent == "Africa"])

gapminder$mean.continent.GDP[gapminder$continent == "Americas"] <- mean(gapminder$gdpPercap[gapminder$continent == "Americas"])

gapminder$mean.continent.GDP[gapminder$continent == "Asia"] <- mean(gapminder$gdpPercap[gapminder$continent == "Asia"])

gapminder$mean.continent.GDP[gapminder$continent == "Europe"] <- mean(gapminder$gdpPercap[gapminder$continent == "Europe"])

gapminder$mean.continent.GDP[gapminder$continent == "Oceania"] <- mean(gapminder$gdpPercap[gapminder$continent == "Oceania"])

You can see how this can get pretty tedious, especially if we want to calculate more complicated or refined statistics. We could use loops or apply functions, but these can be difficult, slow, or error-prone.

dplyr “split-apply-combine”

The abstract problem we’re encountering here is known as split-apply-combine:

We want to split our data into groups (in this case continents), apply some calculations on each group, then combine the results together afterwards.

There are ways to do split-apply-combine operations using the apply() family of functions, but those are error prone and messy.

Luckily, dplyr offers a much cleaner solution to this problem.

# remove the column se just added... There are two easy ways!
gapminder <- gapminder %>% select(-mean.continent.GDP)
# OR
gapminder$mean.continent.GDP <- NULL

dplyr group_by()

We’ve already seen how filter() can help us select observations that meet certain criteria (in the above: continent == "Europe"). More helpful, however, is the group_by() function, which will essentially use every unique criteria that we could have used in filter().

A grouped_df can be thought of as a list where each item in the list is a data.frame which contains only the rows that correspond to the a particular value continent (at least in the example above).

dplyr summarize()

group_by() on its own is not particularly interesting. It’s much more exciting used in conjunction with the summarize() function, which allows use to create new variable(s) by applying transformations to variables in each of the continent-specific data frames.

In other words, when using the group_by() function, we split our original dataframe into multiple pieces, which we then apply summary functions to (e.g. mean() or sd()) within summarize(). The output is a new dataframe reduced in size, with one row per group.

gdp_bycontinents <- gapminder %>%
    group_by(continent) %>%
    summarize(mean_gdpPercap = mean(gdpPercap))
head(gdp_bycontinents)
## # A tibble: 5 x 2
##   continent mean_gdpPercap
##   <fct>              <dbl>
## 1 Africa             2194.
## 2 Americas           7136.
## 3 Asia               7902.
## 4 Europe            14469.
## 5 Oceania           18622.

That allowed us to calculate the mean gdpPercap for each continent. But it gets even better – the function group_by() allows us to group by multiple variables. Let’s group by year and continent.

gdp_bycontinents_byyear <- gapminder %>%
    group_by(continent, year) %>%
    summarize(mean_gdpPercap = mean(gdpPercap))
## `summarise()` has grouped output by 'continent'. You can override using the `.groups` argument.
head(gdp_bycontinents_byyear)
## # A tibble: 6 x 3
## # Groups:   continent [1]
##   continent  year mean_gdpPercap
##   <fct>     <int>          <dbl>
## 1 Africa     1952          1253.
## 2 Africa     1957          1385.
## 3 Africa     1962          1598.
## 4 Africa     1967          2050.
## 5 Africa     1972          2340.
## 6 Africa     1977          2586.

That is already quite powerful, but it gets even better! You’re not limited to defining one new variable in summarize().

gdp_pop_bycontinents_byyear <- gapminder %>%
    group_by(continent, year) %>%
    summarize(mean_gdpPercap = mean(gdpPercap),
              sd_gdpPercap = sd(gdpPercap),
              mean_pop = mean(pop),
              sd_pop = sd(pop))
## `summarise()` has grouped output by 'continent'. You can override using the `.groups` argument.
head(gdp_pop_bycontinents_byyear)
## # A tibble: 6 x 6
## # Groups:   continent [1]
##   continent  year mean_gdpPercap sd_gdpPercap mean_pop    sd_pop
##   <fct>     <int>          <dbl>        <dbl>    <dbl>     <dbl>
## 1 Africa     1952          1253.         983. 4570010.  6317450.
## 2 Africa     1957          1385.        1135. 5093033.  7076042.
## 3 Africa     1962          1598.        1462. 5702247.  7957545.
## 4 Africa     1967          2050.        2848. 6447875.  8985505.
## 5 Africa     1972          2340.        3287. 7305376. 10130833.
## 6 Africa     1977          2586.        4142. 8328097. 11585184.

dplyr mutate()

What if we wanted to extend our original data frame with these values instead of creating a new object? For this, we can use the mutate() function, which is similar to summarize() except it creates new variables to the same dataframe that you pass into it.

gapminder_with_extra_vars <- gapminder %>%
    group_by(continent, year) %>%
    mutate(mean_gdpPercap = mean(gdpPercap),
              sd_gdpPercap = sd(gdpPercap),
              mean_pop = mean(pop),
              sd_pop = sd(pop))
head(gapminder_with_extra_vars)
## # A tibble: 6 x 10
## # Groups:   continent, year [6]
##   country  year    pop continent lifeExp gdpPercap mean_gdpPercap sd_gdpPercap
##   <fct>   <int>  <dbl> <fct>       <dbl>     <dbl>          <dbl>        <dbl>
## 1 Afghan…  1952 8.43e6 Asia         28.8      779.          5195.       18635.
## 2 Afghan…  1957 9.24e6 Asia         30.3      821.          5788.       19507.
## 3 Afghan…  1962 1.03e7 Asia         32.0      853.          5729.       16416.
## 4 Afghan…  1967 1.15e7 Asia         34.0      836.          5971.       14063.
## 5 Afghan…  1972 1.31e7 Asia         36.1      740.          8187.       19088.
## 6 Afghan…  1977 1.49e7 Asia         38.4      786.          7791.       11816.
## # … with 2 more variables: mean_pop <dbl>, sd_pop <dbl>

We can use also use mutate() to create new variables prior to (or even after) summarizing information.

gdp_pop_bycontinents_byyear <- gapminder %>%
    mutate(gdp_billion = gdpPercap*pop/10^9) %>%
    group_by(continent, year) %>%
    summarize(mean_gdpPercap = mean(gdpPercap),
              sd_gdpPercap = sd(gdpPercap),
              mean_pop = mean(pop),
              sd_pop = sd(pop),
              mean_gdp_billion = mean(gdp_billion),
              sd_gdp_billion = sd(gdp_billion))
## `summarise()` has grouped output by 'continent'. You can override using the `.groups` argument.
head(gdp_pop_bycontinents_byyear)
## # A tibble: 6 x 8
## # Groups:   continent [1]
##   continent  year mean_gdpPercap sd_gdpPercap mean_pop sd_pop mean_gdp_billion
##   <fct>     <int>          <dbl>        <dbl>    <dbl>  <dbl>            <dbl>
## 1 Africa     1952          1253.         983. 4570010. 6.32e6             5.99
## 2 Africa     1957          1385.        1135. 5093033. 7.08e6             7.36
## 3 Africa     1962          1598.        1462. 5702247. 7.96e6             8.78
## 4 Africa     1967          2050.        2848. 6447875. 8.99e6            11.4 
## 5 Africa     1972          2340.        3287. 7305376. 1.01e7            15.1 
## 6 Africa     1977          2586.        4142. 8328097. 1.16e7            18.7 
## # … with 1 more variable: sd_gdp_billion <dbl>

dplyr arrange()

As a last step, let’s say we want to sort the rows in our data frame according to values in a certain column. We can use the arrange() function to do this. For instance, let’s organize our rows by year (recent first), and then by continent.

gapminder_with_extra_vars <- gapminder %>%
    group_by(continent, year) %>%
    mutate(mean_gdpPercap = mean(gdpPercap),
              sd_gdpPercap = sd(gdpPercap),
              mean_pop = mean(pop),
              sd_pop = sd(pop)) %>%
    arrange(desc(year), continent)
head(gapminder_with_extra_vars)
## # A tibble: 6 x 10
## # Groups:   continent, year [1]
##   country  year    pop continent lifeExp gdpPercap mean_gdpPercap sd_gdpPercap
##   <fct>   <int>  <dbl> <fct>       <dbl>     <dbl>          <dbl>        <dbl>
## 1 Algeria  2007 3.33e7 Africa       72.3     6223.          3089.        3618.
## 2 Angola   2007 1.24e7 Africa       42.7     4797.          3089.        3618.
## 3 Benin    2007 8.08e6 Africa       56.7     1441.          3089.        3618.
## 4 Botswa…  2007 1.64e6 Africa       50.7    12570.          3089.        3618.
## 5 Burkin…  2007 1.43e7 Africa       52.3     1217.          3089.        3618.
## 6 Burundi  2007 8.39e6 Africa       49.6      430.          3089.        3618.
## # … with 2 more variables: mean_pop <dbl>, sd_pop <dbl>

dplyr Take-aways

# without pipes:

gapminder_with_extra_vars <- arrange(
    mutate(
      group_by(gapminder, continent, year), 
      mean_gdpPercap = mean(gdpPercap)
      ),
    desc(year), continent)

Part II: Tidying Data

Even before we conduct analysis or calculations, we need to put our data into the correct format. The goal here is to rearrange a messy dataset into one that is tidy

The two most important properties of tidy data are:

  1. Each column is a variable.
  2. Each row is an observation.

Tidy data is easier to work with, because you have a consistent way of referring to variables (as column names) and observations (as row indices). It then becomes easy to manipulate, visualize, and model.

For more on the concept of tidy data, read Hadley Wickham’s paper here

Tidying Data: “Wide” vs. “Long” Formats

“Tidy datasets are all alike but every messy dataset is messy in its own way.” – Hadley Wickham

Tabular datasets can be arranged in many ways. For instance, consider the data below. Each data set displays information on heart rate observed in individuals across 3 different time periods. But the data are organized differently in each table.

wide <- data.frame(
  name = c("Wilbur", "Petunia", "Gregory"),
  time1 = c(67, 80, 64),
  time2 = c(56, 90, 50),
  time3 = c(70, 67, 101)
)
wide
##      name time1 time2 time3
## 1  Wilbur    67    56    70
## 2 Petunia    80    90    67
## 3 Gregory    64    50   101
long <- data.frame(
  name = c("Wilbur", "Petunia", "Gregory", "Wilbur", "Petunia", "Gregory", "Wilbur", "Petunia", "Gregory"),
  time = c(1, 1, 1, 2, 2, 2, 3, 3, 3),
  heartrate = c(67, 80, 64, 56, 90, 50, 70, 67, 10)
)
long
##      name time heartrate
## 1  Wilbur    1        67
## 2 Petunia    1        80
## 3 Gregory    1        64
## 4  Wilbur    2        56
## 5 Petunia    2        90
## 6 Gregory    2        50
## 7  Wilbur    3        70
## 8 Petunia    3        67
## 9 Gregory    3        10

Question: Which one of these do you think is the tidy format?

Answer: The first dataframe (the “wide” one) would not be considered tidy because values (i.e., heartrate) are spread across multiple columns.

We often refer to these different structurs as “long” vs. “wide” formats. In the “long” format, you usually have 1 column for the observed variable and the other columns are ID variables.

For the “wide” format each row is often a site/subject/patient and you have multiple observation variables containing the same type of data. These can be either repeated observations over time, or observation of multiple variables (or a mix of both). In the above case, we had the same kind of data (heart rate) entered across 3 different columns, corresponding to three different time periods.

You may find data input may be simpler and some programs/functions may prefer the “wide” format. However, many of R’s functions have been designed assuming you have “long” format data.

Tidying the gapminder Data

Let’s revisit the structure of our gapminder dataframe:

head(gapminder)
##       country year      pop continent lifeExp gdpPercap
## 1 Afghanistan 1952  8425333      Asia  28.801  779.4453
## 2 Afghanistan 1957  9240934      Asia  30.332  820.8530
## 3 Afghanistan 1962 10267083      Asia  31.997  853.1007
## 4 Afghanistan 1967 11537966      Asia  34.020  836.1971
## 5 Afghanistan 1972 13079460      Asia  36.088  739.9811
## 6 Afghanistan 1977 14880372      Asia  38.438  786.1134

Question: Is this data frame wide or long?

Answer: This data frame is somewhere in between the purely ‘long’ and ‘wide’ formats:

Despite not having ALL observations in one column, this intermediate format makes sense given that all three observation variables have different units. As we have seen, many of the functions in R are often vector based, and you usually do not want to do mathematical operations on values with different units.

On the other hand, there are some instances in which a purely long or wide format is ideal (e.g. plotting). Likewise, sometimes you’ll get data on your desk that is poorly organized, and you’ll need to reshape it.

Enter the tidyr package…

Thankfully, the tidyr package will help you efficiently transform your data regardless of original format.

# Install the "tidyr" package 
# install.packages("tidyr") # Not Run

# Load the "tidyr" package (necessary every new R session)
library(tidyr)

tidyr gather()

Until now, we’ve been using the nicely formatted original gapminder dataset. This dataset is not quite wide and not quite long – it’s something in the middle, but ‘real’ data (i.e. our own research data) will never be so well organized. Here let’s start with the wide format version of the gapminder dataset.

gap_wide <- read.csv("../data/gapminder_wide.csv", stringsAsFactors = FALSE)
head(gap_wide)
##   continent      country gdpPercap_1952 gdpPercap_1957 gdpPercap_1962
## 1    Africa      Algeria      2449.0082      3013.9760      2550.8169
## 2    Africa       Angola      3520.6103      3827.9405      4269.2767
## 3    Africa        Benin      1062.7522       959.6011       949.4991
## 4    Africa     Botswana       851.2411       918.2325       983.6540
## 5    Africa Burkina Faso       543.2552       617.1835       722.5120
## 6    Africa      Burundi       339.2965       379.5646       355.2032
##   gdpPercap_1967 gdpPercap_1972 gdpPercap_1977 gdpPercap_1982 gdpPercap_1987
## 1      3246.9918      4182.6638      4910.4168      5745.1602      5681.3585
## 2      5522.7764      5473.2880      3008.6474      2756.9537      2430.2083
## 3      1035.8314      1085.7969      1029.1613      1277.8976      1225.8560
## 4      1214.7093      2263.6111      3214.8578      4551.1421      6205.8839
## 5       794.8266       854.7360       743.3870       807.1986       912.0631
## 6       412.9775       464.0995       556.1033       559.6032       621.8188
##   gdpPercap_1992 gdpPercap_1997 gdpPercap_2002 gdpPercap_2007 lifeExp_1952
## 1      5023.2166      4797.2951      5288.0404      6223.3675       43.077
## 2      2627.8457      2277.1409      2773.2873      4797.2313       30.015
## 3      1191.2077      1232.9753      1372.8779      1441.2849       38.223
## 4      7954.1116      8647.1423     11003.6051     12569.8518       47.622
## 5       931.7528       946.2950      1037.6452      1217.0330       31.975
## 6       631.6999       463.1151       446.4035       430.0707       39.031
##   lifeExp_1957 lifeExp_1962 lifeExp_1967 lifeExp_1972 lifeExp_1977 lifeExp_1982
## 1       45.685       48.303       51.407       54.518       58.014       61.368
## 2       31.999       34.000       35.985       37.928       39.483       39.942
## 3       40.358       42.618       44.885       47.014       49.190       50.904
## 4       49.618       51.520       53.298       56.024       59.319       61.484
## 5       34.906       37.814       40.697       43.591       46.137       48.122
## 6       40.533       42.045       43.548       44.057       45.910       47.471
##   lifeExp_1987 lifeExp_1992 lifeExp_1997 lifeExp_2002 lifeExp_2007 pop_1952
## 1       65.799       67.744       69.152       70.994       72.301  9279525
## 2       39.906       40.647       40.963       41.003       42.731  4232095
## 3       52.337       53.919       54.777       54.406       56.728  1738315
## 4       63.622       62.745       52.556       46.634       50.728   442308
## 5       49.557       50.260       50.324       50.650       52.295  4469979
## 6       48.211       44.736       45.326       47.360       49.580  2445618
##   pop_1957 pop_1962 pop_1967 pop_1972 pop_1977 pop_1982 pop_1987 pop_1992
## 1 10270856 11000948 12760499 14760787 17152804 20033753 23254956 26298373
## 2  4561361  4826015  5247469  5894858  6162675  7016384  7874230  8735988
## 3  1925173  2151895  2427334  2761407  3168267  3641603  4243788  4981671
## 4   474639   512764   553541   619351   781472   970347  1151184  1342614
## 5  4713416  4919632  5127935  5433886  5889574  6634596  7586551  8878303
## 6  2667518  2961915  3330989  3529983  3834415  4580410  5126023  5809236
##   pop_1997 pop_2002 pop_2007
## 1 29072015 31287142 33333216
## 2  9875024 10866106 12420476
## 3  6066080  7026113  8078314
## 4  1536536  1630347  1639131
## 5 10352843 12251209 14326203
## 6  6121610  7021078  8390505

The first step towards getting our nice intermediate data format is to first convert from the wide to the long format. The function gather() will ‘gather’ the observation variables into a single variable. This is sometimes called “melting” your data, because it melts the table from wide to long. Those data will be melted into two variables: one for the variable names, and the other for the variable values.

gap_long <- gap_wide %>%
    gather(obstype_year, obs_values, 3:38)
head(gap_long)
##   continent      country   obstype_year obs_values
## 1    Africa      Algeria gdpPercap_1952  2449.0082
## 2    Africa       Angola gdpPercap_1952  3520.6103
## 3    Africa        Benin gdpPercap_1952  1062.7522
## 4    Africa     Botswana gdpPercap_1952   851.2411
## 5    Africa Burkina Faso gdpPercap_1952   543.2552
## 6    Africa      Burundi gdpPercap_1952   339.2965

Notice that we put three arguments into the gather() function:

  1. the name for the new column for the new ID variable (obstype_year),
  2. the name for the new amalgamated observation variable (obs_value),
  3. the indices of the old variables (3:38, signalling columns 3 through 38) that we want to gather into one variable. Notice that we don’t want to melt down columns 1 and 2; these are considered “ID” variables.

tidyr select()

If there are many columns or they’re named in a consistent pattern, we might not want to select them using the column numbers. Sometimes it’s easier to use some information contained in the names themselves.

We can select variables using:

See the select() function in dplyr for more options.

For instance, here we do the same gather operation with (1) the starts_with function, and (2) the - operator:

# with the starts_with() function
gap_long <- gap_wide %>%
    gather(obstype_year, obs_values, starts_with('pop'),
           starts_with('lifeExp'), starts_with('gdpPercap'))
head(gap_long)
##   continent      country obstype_year obs_values
## 1    Africa      Algeria     pop_1952    9279525
## 2    Africa       Angola     pop_1952    4232095
## 3    Africa        Benin     pop_1952    1738315
## 4    Africa     Botswana     pop_1952     442308
## 5    Africa Burkina Faso     pop_1952    4469979
## 6    Africa      Burundi     pop_1952    2445618
# with the - operator
gap_long <- gap_wide %>% 
  gather(obstype_year, obs_values, -continent, -country)
head(gap_long)
##   continent      country   obstype_year obs_values
## 1    Africa      Algeria gdpPercap_1952  2449.0082
## 2    Africa       Angola gdpPercap_1952  3520.6103
## 3    Africa        Benin gdpPercap_1952  1062.7522
## 4    Africa     Botswana gdpPercap_1952   851.2411
## 5    Africa Burkina Faso gdpPercap_1952   543.2552
## 6    Africa      Burundi gdpPercap_1952   339.2965

However you choose to do it, notice that the output collapses all of the measure variables into two columns: one containing new ID variable, the other containing the observation value for that row.

tidyr separate()

You’ll notice that in our long dataset, obstype_year actually contains 2 pieces of information, the observation type (pop, lifeExp, or gdpPercap) and the year.

We can use the separate() function to split the character strings into multiple variables:

gap_long_sep <- gap_long %>% 
  separate(obstype_year, into = c('obs_type','year'), sep = "_") %>% 
  mutate(year = as.integer(year))
head(gap_long_sep)
##   continent      country  obs_type year obs_values
## 1    Africa      Algeria gdpPercap 1952  2449.0082
## 2    Africa       Angola gdpPercap 1952  3520.6103
## 3    Africa        Benin gdpPercap 1952  1062.7522
## 4    Africa     Botswana gdpPercap 1952   851.2411
## 5    Africa Burkina Faso gdpPercap 1952   543.2552
## 6    Africa      Burundi gdpPercap 1952   339.2965

If you didn’t use tidyr to do this, you’d have to use the strsplit function and use multiple lines of code to replace the column in gap_long with two new columns. This solution is much cleaner.

tidyr spread

The opposite of gather() is spread(). It spreads our observation variables back out to make a wider table. We can use this function to spread our gap_long() to the original “medium” format.

gap_medium <- gap_long_sep %>% 
  spread(obs_type, obs_values)
head(gap_medium)
##   continent country year gdpPercap lifeExp      pop
## 1    Africa Algeria 1952  2449.008  43.077  9279525
## 2    Africa Algeria 1957  3013.976  45.685 10270856
## 3    Africa Algeria 1962  2550.817  48.303 11000948
## 4    Africa Algeria 1967  3246.992  51.407 12760499
## 5    Africa Algeria 1972  4182.664  54.518 14760787
## 6    Africa Algeria 1977  4910.417  58.014 17152804

All we need is some quick fixes to make this dataset identical to the original gapminder dataset:

gapminder <- read.csv("../data/gapminder-FiveYearData.csv")
head(gap_medium)
##   continent country year gdpPercap lifeExp      pop
## 1    Africa Algeria 1952  2449.008  43.077  9279525
## 2    Africa Algeria 1957  3013.976  45.685 10270856
## 3    Africa Algeria 1962  2550.817  48.303 11000948
## 4    Africa Algeria 1967  3246.992  51.407 12760499
## 5    Africa Algeria 1972  4182.664  54.518 14760787
## 6    Africa Algeria 1977  4910.417  58.014 17152804
head(gapminder)
##       country year      pop continent lifeExp gdpPercap
## 1 Afghanistan 1952  8425333      Asia  28.801  779.4453
## 2 Afghanistan 1957  9240934      Asia  30.332  820.8530
## 3 Afghanistan 1962 10267083      Asia  31.997  853.1007
## 4 Afghanistan 1967 11537966      Asia  34.020  836.1971
## 5 Afghanistan 1972 13079460      Asia  36.088  739.9811
## 6 Afghanistan 1977 14880372      Asia  38.438  786.1134
# rearrange columns
gap_medium <- gap_medium[,names(gapminder)]
head(gap_medium)
##   country year      pop continent lifeExp gdpPercap
## 1 Algeria 1952  9279525    Africa  43.077  2449.008
## 2 Algeria 1957 10270856    Africa  45.685  3013.976
## 3 Algeria 1962 11000948    Africa  48.303  2550.817
## 4 Algeria 1967 12760499    Africa  51.407  3246.992
## 5 Algeria 1972 14760787    Africa  54.518  4182.664
## 6 Algeria 1977 17152804    Africa  58.014  4910.417
# arrange by country, continent, and year
gap_medium <- gap_medium %>% 
  arrange(country,continent,year)
head(gap_medium)
##       country year      pop continent lifeExp gdpPercap
## 1 Afghanistan 1952  8425333      Asia  28.801  779.4453
## 2 Afghanistan 1957  9240934      Asia  30.332  820.8530
## 3 Afghanistan 1962 10267083      Asia  31.997  853.1007
## 4 Afghanistan 1967 11537966      Asia  34.020  836.1971
## 5 Afghanistan 1972 13079460      Asia  36.088  739.9811
## 6 Afghanistan 1977 14880372      Asia  38.438  786.1134

Extra Resources

dplyr and tidyr have many more functions to help you wrangle and manipulate your data. See the Data Wrangling Cheat Sheet for more.

There are some other useful packages in the tidyverse:

Exercises:

dplyr

  1. Use dplyr to create a data frame containing the median lifeExp for each continent

  2. Use dplyr to add a column to the gapminder dataset that contains the total population of the continent of each observation in a given year. For example, if the first observation is Afghanistan in 1952, the new column would contain the population of Asia in 1952.

  3. Use dplyr to: add a column called gdpPercap_diff that contains the difference between the observation’s gdpPercap and the mean gdpPercap of the continent in that year. Arrange the dataframe by the column you just created, in descending order (so that the relatively richest country/years are listed first)

hint: You might have to ungroup() before you arrange().

tidyr

  1. Subset the results from question #3 to select only the country, year, and gdpPercap_diff columns. Use tidyr put it in wide format so that countries are rows and years are columns.

Next Module